;
};
TIN adds [the `==>` operator](/docs/postgres/search/reference/operator). The query on the right-hand side of it is **TINQL**.
```sql theme={null}
SELECT id, body
FROM posts
WHERE body ==> 'apple AND "fuji apple"';
```
**Rules**
* Keywords must be **UPPER CASE** (`AND`, `OR`, `THEN`). Lowercase is a search term (`and` matches the word “and”).
* Matching **folds case and accents** under the default tokenizer (`Jalapeño` and `jalapeno` are the same term). Override with `case_folding` / `accent_folding` on the index.
* Inputs that analyze to **no tokens** (`''`, whitespace, bare punctuation) match nothing. Explicit empty syntax (`""`, `[]`) is a parse error.
* Query boost factors (`^N`) must be in `[0.0..10000.0]`. The system rejects out-of-range values at parse time.
* Token positions are **0-based**. Prefer `IN FIRST N WORDS` / `IN LAST N WORDS` over raw `IN WORDS` ranges when you mean “the first/last N words.”
* Juxtaposition defaults to AND: `apple grape` acts as `apple AND grape`.
* Exclude terms with `AND NOT`: `apple AND NOT peel`.
* Quote reserved words used as terms: `"AND"`, `"THEN"`.
* A hyphen usually splits a term: `wi-fi` is searched as `"wi fi"`.
* Fuzzy match (`apple~2`) works on one word. `wi-fi~2` errors because `wi-fi` is two words.
## Terms
| Form | Meaning | Example |
| ------------------- | --------------------------------------------------------- | ---------------------------------------------------------- |
| bare term | Single dictionary term (case + accent folded by default) | `apple` (same as `Apple`), `jalapeno` (same as `Jalapeño`) |
| `*` | Match-all documents (standalone only) | `*` |
| `*` / `?` wildcards | `*` = 0+ chars, `?` = 1 char | `appl*`, `p?ach`, `*house` |
| `term~N` | Fuzzy: edit distance N, stable prefix 1 | `apple~2` |
| `term~P:N` | Fuzzy: stable prefix P, distance N | `apple~0:2` |
| `A TO B` | Inclusive term-dictionary range | `aardvark TO cat` |
| `* TO B` / `A TO *` | Open range bound | `* TO cat` |
| `MATCHES regex` | Full-term regex (to unescaped whitespace); **not** folded | `MATCHES peach.*` |
| `CONTAINS term` | Same as bare term (optional) | `CONTAINS apple` |
| `expr^N` | Boost (`N` ∈ `[0.0..10000.0]`) | `apple^2`, `"x"^1.5` |
Allowed in terms: anything except whitespace and `( ) [ ] " ~ ^`. Escape literal `*` / `?` with `\`. Space in a regex: `MATCHES foo\ bar`.
Because `MATCHES` is not folded, write patterns against the normalized dictionary form (usually lowercase, accents folded): `MATCHES apple.*`, not `MATCHES Apple.*`.
## Phrases
| Form | Meaning | Example |
| ------------- | ------------------------------------------ | ------------------------ |
| `"a b c"` | Adjacent ordered words | `"fuji apple"` |
| `"a _ c"` | `_` = one any-word gap | `"big _ wolf"` |
| `"a [b c] d"` | Per-position alternatives | `"big [bad large] wolf"` |
| `"a b"~N` | Up to N extra word gaps (phrase tolerance) | `"fuji apple"~2` |
Phrase escapes: `\"` `\\` `\_` `\[` `\]`. Empty `""` is invalid (parse error). Inputs that analyze to no tokens match nothing.
## Boolean operators
| Operator | Meaning | Example |
| ------------- | ------------------------ | -------------------- |
| `A AND B` | Both match | `juicy AND apple` |
| `A OR B` | Either matches | `apple OR grape` |
| `A AND NOT B` | A matches and B does not | `apple AND NOT peel` |
| `A B` | Implicit AND | `apple grape` |
`NOT` never stands alone. It only appears in `AND NOT`, `NOT ENCLOSES`, `NOT ENCLOSED BY`, and `NOT OVERLAPPING`.
## Alternatives
| Form | Meaning | Example |
| -------------------- | -------------------------------------------- | --------------------------- |
| `[a b c]` | OR of whitespace-separated alts (not commas) | `[mango plum pear]` |
| `AT LEAST N OF […]` | At least N alts match | `AT LEAST 2 OF [a b c d]` |
| `AT LEAST N% OF […]` | At least N% of alts match | `AT LEAST 50% OF [a b c d]` |
| `ALL OF […]` | Every alt matches | `ALL OF [a b c]` |
Alternatives may hold expressions: `["big bad" house NEAR/2 brick]`.
## Proximity
`/N` is required on `THEN` and `NEAR`. N is the maximum number of extra words allowed between the operands, and `0` means adjacent.
| Operator | Meaning | Example |
| ------------------ | -------------------- | -------------------------------- |
| `A THEN/N B` | Ordered proximity | `fuji THEN/0 apple` |
| `A NEAR/N B` | Either order | `peach NEAR/5 blossom` |
| `(A … B) WITHIN N` | Match span width ≤ N | `(citrus NEAR/5 melon) WITHIN 6` |
## Span relations
Every TINQL expression produces **spans** (position ranges). Relation operators compare two span sets.
| Operator | Keeps | Example |
| --------------------- | ----------------------------------------- | ------------------------------------------------------ |
| `A ENCLOSES B` | Spans of A that contain a span of B | `(security NEAR/10 threat) ENCLOSES critical` |
| `A NOT ENCLOSES B` | Spans of A that contain no span of B | `* NOT ENCLOSES spam` |
| `A ENCLOSED BY B` | Spans of A that fall inside a span of B | `critical ENCLOSED BY (security NEAR/10 threat)` |
| `A NOT ENCLOSED BY B` | Spans of A not inside any span of B | `price NOT ENCLOSED BY (disclaimer NEAR/20 terms)` |
| `A OVERLAPPING B` | Spans of A that share a position with B | `(apple NEAR/5 fuji) OVERLAPPING (kiwi NEAR/5 citrus)` |
| `A NOT OVERLAPPING B` | Spans of A that share no positions with B | `title NOT OVERLAPPING disclaimer` |
| `A BEFORE B` | Spans of A that start before B | `abstract BEFORE conclusion` |
| `A AFTER B` | Spans of A that start after B | `price AFTER discount` |
`ENCLOSES` and `ENCLOSED BY` test the same relationship but return different spans: `ENCLOSES` returns the outer span and `ENCLOSED BY` returns the inner one.
## Positional filters
Token positions are **0-based** (the first token is position `0`). Use `IN FIRST` / `IN LAST` when you care about word counts, and `IN WORDS` when you need an explicit position window.
| Form | Meaning | Example |
| -------------------- | --------------------------------------------------- | ---------------------------- |
| `A IN FIRST N WORDS` | Match in the first N tokens (positions `0` … `N-1`) | `apple IN FIRST 100 WORDS` |
| `A IN FIRST N%` | Match in first N% of the document | `apple IN FIRST 25%` |
| `A IN LAST N WORDS` | Match in the last N tokens | `apple IN LAST 50 WORDS` |
| `A IN LAST N%` | Match in last N% of the document | `apple IN LAST 25%` |
| `A IN MIDDLE N%` | Match in middle N% of the document | `apple IN MIDDLE 50%` |
| `A IN WORDS X TO Y` | Match in inclusive 0-based position range `[X, Y]` | `apple IN WORDS 500 TO 1000` |
## Precedence (loose → tight)
| Level | Operators |
| ----- | --------------------------------------------------- |
| 1 | `OR` |
| 2 | `AND` (explicit and implicit) |
| 3 | `AND NOT` |
| 4 | Positional filters (`IN …`) |
| 5 | Relation operators (`ENCLOSES`, `BEFORE`, …) |
| 6 | Proximity (`THEN/N`, `NEAR/N`) |
| 7 | `WITHIN` |
| 8 | Boost (`^N`) |
| 9 | Primary (`apple`, `"phrase"`, `[a b]`, `MATCHES …`) |
Operators at the same level are left-associative. Use parentheses to override precedence.
`A OR B THEN/5 C AND D` → `A OR ((B THEN/5 C) AND D)`.
Postfix modifiers compose on the primary: `apple~2^3`, `"big bad"~2^1.5`.
## Keyword index
All keywords are case-sensitive **UPPER CASE**.
| Category | Keywords |
| ---------------- | -------------------------------------------------------------- |
| Boolean | `AND`, `OR`, `NOT` (only in compounds) |
| Proximity | `THEN`, `NEAR`, `WITHIN` |
| Relations | `ENCLOSES`, `ENCLOSED`, `BY`, `OVERLAPPING`, `BEFORE`, `AFTER` |
| Range / position | `TO`, `IN`, `FIRST`, `LAST`, `MIDDLE`, `WORDS` |
| Terms | `CONTAINS`, `MATCHES` |
| Alternatives | `AT`, `LEAST`, `OF`, `ALL` |
`IN` is special only before `FIRST` / `LAST` / `MIDDLE` / `WORDS`. `AT` only before `LEAST`. `ALL` only before `OF`. `ENCLOSED` / `BY` only together.
## Special characters
| Character | Meaning |
| --------- | -------------------------------------------------------- |
| `"…"` | Phrase |
| `(…)` | Grouping |
| `[…]` | Alternatives |
| `~N` | Fuzzy (on term) or phrase word-gap tolerance (after `"`) |
| `~P:N` | Fuzzy with explicit prefix length |
| `^N` | Boost |
| `*` | Wildcard, open range bound, or standalone match-all |
| `?` | Single-character wildcard |
| `_` | One-word gap inside a phrase |
| `\` | Escape |
## Need help?
Get help from [the PlanetScale Support team](https://planetscale.com/contact?initial=support), or join our [Discord community](https://pscale.link/community) to see how others are using PlanetScale.